The question is a reasonable one, and I'd think your position is the correct one, i.e. while a certain box might win more races the odds available over time are reduced to the extent that backing all starters from that box is not a winning strategy.
To draw any valid conclusions we need a sizeable dataset, otherwise we're wasting our time.
I have a lot of data from the 1980s. You might not be interested in it, but it does address the key issues.
I'll post just one table for now - I have many others, with breakdowns by price brackets, and profitability, etc.
The data comes from the Gabba, where we had a pretty strong betting ring. Not sure where you'd get that these days.
The Gabba was a crazy track, but the most-used start (558m) was reasonably fair, starting from the top of the straight. I'd say the circuit was about 450m in circumference.
I got results for 900 consecutive races with full fields. Love that dot matrix printer!!!
We always worked on the rule that the red box was the best. The data confirms that and without doubt the finding is statistically significant. The 8 was next best with 2, 3 and 7 a bit behind.
Not surprisingly, the bias shows us to a similar degree in the PLACE results.
A very informative statistic is the % Opposition Beaten. I've written about this before, but no one that I know of uses it. It really is the only meaningful measure of the advantage (or otherwise) of a box (or barrier).
Again, the red dogs beat hom almost 60% of their opposition. This compares to an expected 50% - if all was fair. We tend to shy away from the 4 and 5, but you can see that their respective stats are 49 and 46 - probably not statistically significant versus 50%.
Now, what about odds? There's a column for the SP average. Apart from the red box (at 6/1) the others are reasonably close between 10/1 and 13/1. Average can be a misleading statistic - frrequently misused - so we should treat it with caution.
The two columns on the far right are interesting.
Note that while the red box wins 21% of the time, and we average 6/1 of all runners, we still lose $226 when we have $1 win on all runners. (Note my comment above re average.) This means that the longer-priced runners do not perform to their price (probability). In fact, the best result for this approach comes from box 7.
The very far right column is the most meaningful. This shows the result when we back every runner to RETURN $100. That is, if the dog was 4/1, we would have $20 on it. Note that again box 7 is the best outcome - it doesn't have to be that way just because the $1 WIN has the best outcome - while the red box is pretty close to 4, 5, and 8.
The value of the RETURN $100 stat is that it tells us whether the short-priced runners perform to expectations. From this table, we can conclude that when a 7 dog is at short odds it will perform well. On the other hand, from box 2 and 6, the favourites went poorly.
Hope this is of some interest.
Happy to discuss further if you wish.